Analysis of Proportional Fair Scheduling Algorithm for...

International Journal of Control and Automation

Vol. 11, No. 12 (2018), pp.11-24

http//dx.doi.org/10.14257/ijca.2018.11.12.02

ISSN: 2005-4297 IJCA

Copyright © 2018 SERSC Australia

Analysis of Proportional Fair Scheduling Algorithm for Long

Term Evolution (LTE) Uplink

1

Endah Budi Purnomowati1, Muthia Rahma2 and Rudy Yuwono3

1,2,3Dept. of Electrical Engineering, Brawijaya University

MT. Haryono 167, Malang 65145, Indonesia 2Telecommunication Labotatory, Brawijaya University

MT. Haryono 167, Malang 65145, Indonesia 3Transmission and Microwave Laboratory, Brawijaya University

MT. Haryono 167, Malang 65145, Indonesia [email protected], [email protected], [email protected]

Abstract

Scheduling is a radio resource allocation schemes for serving users by an algorithm.

The amount of resource blocks that will be allocated is 50 by the 10 MHz bandwidth. The

simulation scenario defined as a single cell network with several users in which located

among 1-4 km. Proportional Fair algorithm will assigns more resource block to a user who

has the highest rate than the average rate. This simulation result shows that SNR value

implicates the data rate and modulation scheme applied. The increasing of user distance

from eNodeB, will decrease SNR and data rate. Thus, the number of resource block

allocated will be decreased. While as the increasing of user number, the value of system’s

BER are tend to constant, maximum normalized throughput are increase, and the level of

fairness will tend to constant close to ideal value of 1.

Keywords: scheduling, Proportional Fair, BER, throughput, fairness

1. Introduction

The enhancement of LTE (Long Term Evolution) user demands efficient methods for

serving a sufficient capacity and good fairness index. Scheduling methods are introduced

to provide the efficiency using resource block allocation algorithm. There are several kind

of algorithm to do scheduling or assign or allocate those resource blocks. Some resource

blocks on certain bandwidth are allocated for active users at each channel. In the LTE uplink

channel, requires Single Carrier Frequency Division Multiple Access (SC-FDMA) that

users cannot be allocated on separated resource blocks because of its single carrier property.

This is the main characteristic of uplink channel that differentiate with the downlink.

In this paper, an analysis of scheduling simulation will be performed in the LTE uplink

channel. The scenarios are set on a single cell network with one eNodeB and some active

users. The amount of users in scenarios A, B, C, and D are 4, 8, 12, and 16 respectively.

The users of each scenarios are located between 1 to 4 kilometers from eNodeB. This user

number variation will obtain quite significant parameter value that can be analyzed as

showed in [12]. The 50 resource blocks will occupy 10 MHz bandwidth for each time slot

(0.5 milliseconds). The scheduling of Proportional Fair algorithm will serve the user with

better channel quality first. While the other users whose qualities are lower will have

minimum scheduling or less allocation of resource blocks. The Proportional Fair algorithm

was chosen because according to [3], it provides balance between fairness index and the

system throughput.

Received (September 16, 2018), Review Result (November 25, 2018), Accepted (December 6, 2018)


Vol. 11, No. 12 (2018)

12 Copyright © 2018 SERSC Australia

The research questions in this paper, we will see how the channel quality on Line of

Sight (LOS) and Non Line of Sight (NLOS) determine the scheduling of resource blocks

based on the average quality of each scenarios. The channel quality shown by the Signal to

Noise Ratio (SNR) value determines the modulation used i.e. QPSK and 16QAM. While

the other parameters to be analyzed are Bit Error Rate (BER), throughput, and fairness.

2. Related Researches

2.1. Scheduling Algorithm

Packet scheduler assigns the number of RB for active users. In SC-FDMA, this

assignment is done on condition that the RB assigned to the user must be sequential on the

frequency domain. The user’s transmission power is equally distributed on each assigned

RB [1].

The scheduling algorithm can be described as an algorithm that assigns the RB to user

according to some parameters or metrics to obtain a good system performance under certain

conditions [2]. There are some type of scheduling algorithm can be used in LTE system

those have different proporties. Proportional Fair is an approach that selects user with the

best channel level compared to the average channel level. This system supports fairness by

prioritizing channels with higher relative quality. A Proportinonal Fair scheduling

algorithm provides balance between fairness and the system throughput. It was first

presented in Code-Division Multiple Access High Data Rates (CDMA-HDR) [3], but is

now used in OFDMA and SC-FDMA based systems as well. Proportional Fair algorithm

will choose user-i and can be defined by Equation (1):

𝑢𝑠𝑒𝑟(𝑖) = 𝑚𝑎𝑥𝑟𝑎𝑡𝑒(𝑖)

𝑟𝑎𝑡𝑒´ .................................................................................................... (1)

2.2 Resource Block

In a 3GPP LTE, slot structure is defined as a group of resource block (RB) that occupies

180kHz bandwidth at frequency domain and 0.5 milliseconds time slot at time domain. A

RB has 12 subcarriers that each subcarriers occupy 15 kHz at frequency domain, and 7

symbols on 0.5 milliseconds at time domain (Figure 1). While 1 Transmission Time

Interval (TTI) consist of two RBs, which TTI is 1 milliseconds in time domain [3]. The

initial channel quality determination until scheduling process will be done per TTI. RB is a

basic element for scheduling in a LTE system with various applicable scheduling methods

to allocate a group of RB to serve active users. In a SC-FDMA, RBs can not be allocated

separately because its single carrier property [4].


Vol. 11, No. 12 (2018)

Copyright © 2018 SERSC Australia 13

Figure 1. LTE Resource Block Model

The smallest unit in LTE scheduling is an intersection of 180kHz with 1 milliseconds

TTI or can be describe as two RBs in a time domain. Scheduling control by eNodeB based

on channel quality of user.

3. Scheduling Simulation

The system design is a LTE single cell network with various user number those are 4, 8,

12, and 16 that will be located in various distance with the eNodeB. The specified channel

type is a Rayleigh Fading propagation channel where there is a pathloss (LOS and NLOS)

caused by distance and multipath fading effect. Figure 2 shows the simple uplink

communication between user and eNodeB to determine the resource allocation. SNR

estimation of each user channel will be used to determine the channel condition. Then, the

scheduling steps are marked by a red arrow.

Figure 2. Uplink Communication Scheme

Among the initial parameters, the propagation losses and the SNR will be related to the

type of modulation and channel capacity. SNR at some level will determine the type of

modulation used where the modulation number will determine BER. While the channel

capacity will be used as reference of scheduling or allocation of radio sources. While the

simulation is done by the following steps:


Vol. 11, No. 12 (2018)


3.1 Check the Channel Conditions

It is known in wireless celular systems that users located in different distances have

different fading condition. Consequently, users that have high SNR leads to unfair

scheduling amongst users [6]. The simulation starts with checking the channel conditions

by calculate the SNR of each user with Equation (2) [7] and the user channel capacity with

Equation (3). Channel conditions obtained from the calculation are mapped in the matrix

(Figure 3) where C represents channel, m represents time, and i represents user number.

This matrix contain different values of channel capacity or data rate and will be processed

at allocation or scheduling for each time.

𝑆𝑁𝑅 = 10𝑙𝑜𝑔𝑃𝑟

𝑁0 ..................................................................................................... (2)

SNR = signal to noise ratio (dB)

Pr = receiver power (mW)

N0 = noise power

𝐶 = 𝐵 × 𝑙𝑜𝑔2(1 + 𝑆𝑁𝑅) ......................................................................................... (3)

C = channel capacity

B = Bandwidth (Hz)

Figure 3. Matrix of Channel Condition

3.2 Scheduling

Scheduling based on Proportional Fair algorithm described in Figure 4 where m is a

time, n is a number of resource blocks, and i is a number of user. Proportional Fair assigns

the resource blocks to user by selecting the best channel level compared to the average

channel level of users. User with the best channel condition will be served by more resource

blocks.

Figure 4. Scheduling Process

3.3 BER

BER is a parameter of channel quality received in a digital transmission systems [8].

The result of BER calculation is converted in a matriks form for each user then summed up

for 1 TTI and divided by the number of system RB, so the average of system BER for each

scenarios are obtained.

Scheduling


Vol. 11, No. 12 (2018)


3.4 Throughput

Throughput of users and throughput of the cell area are affected by the methodology

selected by the scheduling algorithm[9]. Throughput calculation is a calculation of

normalized throughput expressed in Equation (4) [10]. Normalized throughput represents

the value of success which requested by users due to the system bandwidth and efficiency.

In this case, the system bandwidth simply considered as a resource blocks. Based on the

value of throughput normalized obtained, the highest value is selected as a best value of the

system.

𝑁𝑖 =𝐷𝑖

1

𝐾∑ 𝐷𝑖𝐾𝑖=1

........................................................................................................... (4)

Ni = Normalized throughput each user

Di = Data rate each user

K = user number

3.5 Fairness

The values of fairness are between 0 and 1 or 0% until 100%. Fairness for each scenarios

calculated by the Equation (5) [11].

∑ 𝐷𝑖∨2𝐾

𝑖=1

𝐾∙∑ 𝐷𝑖2𝐾

𝑖=1

𝐹 =

................................................................................................................. (5)

F = Fairness index

Di = Data rate each user

K = user number

Simulation and parameter calculations are using Matlab R2015a under the following

characteristics in Table 1.

Table 1. Simulation Parameters

Parameters Value

System Single cell

Carrier frequency (MHz) 700

Bandwidth (MHz) 10

Efficiency (%) 90

Subcarrier 600

Subcarrier spacing (KHz) 15

RB 50

TTI (ms) 1

Transmitter power (dBm) 43

Antenna gain (dB) 18

User number 4, 8, 12, 16

Modulation QPSK dan 16 QAM

User distance (Km) 1-4

4. Experiments

4.1 SNR

Path loss or propagation loss will determine the SNR value for the user is inversely

proportional, which means the higher the path loss the lower the SNR. Better SNRs are of

higher value. SNR LOS condition is higher than NLOS condition at same distance and will


Vol. 11, No. 12 (2018)


decrease with increasing distance between user with eNodeB. This is because, in NLOS

experience more decrease signal level due to multipath.

Both LOS and NLOS condition determine the pathloss value that implicate to SNR

calculation. The carrier frequency is selected at 700 MHz with FDD. SNR value as a

reference is considered since the calculation, scheduling until the final step of performance

calculation. The difference SNR value of the user on the various distance with eNodeB is

shown in Figure 5. Generally, the LOS condition has a higher SNR than the NLOS

condition, which means that users in LOS condition has a higher acceptability and noise

resistance than NLOS with the same distance. This is due to LOS condition which has a

slight of fading effect so the decreasing of power level are less than NLOS condition. Both

in LOS and NLOS condition, the SNR value will be desrease as the increasing of distance

because of error cosidered as noise and attenuation influence. According to the equation

and calculation, SNR will greater when the noise power was lower. And this is implicates

to modulation scheme used by the user and BER which explain in Section 4.4.

Figure 5. SNR to Distance

Based on the SNR calculation before, the SNR value of each user of scenarios can be

determined as in Table 2.

Table 2. SNR User

Scenario User User number SNR (dB)

LOS NLOS

A 4

1 64,4874 44,5932

2, 3 56,8832 32,4265

4 52,4462 25,3273

B 8

1, 2 64,4874 44,5932

3, 4, 5 58,4668 34,9603

6, 7 55,5442 30,2842

8 52,4462 25,3273

C 12

1, 2, 3 64,4874 44,5932

4, 5, 6 59,3819 36,4245

7, 8, 9 55,5442 30,2842

10, 11, 12 52,4462 25,3273

D 16

1, 2, 3, 4 64,4874 44,5932

5, 6 60,4050 38,0614

7, 8, 9, 10 55,5442 30,2842

11, 12 54,3844 28,4284


Vol. 11, No. 12 (2018)


Scenario User User number SNR (dB)

LOS NLOS

13,14, 15, 16 52,4462 25,3273

Source: Design

4.2 Channel Capacity

SNR brings implication to the value of channel capacity in a proportional way. User with

high channel capacity value will have good condition because of high SNR value. Figure 6

shows the channel capacity of users for each scenario in a LOS and NLOS condition. Users

in a NLOS condition have lower channel capacity than the LOS condition with difference

between 30-50% and this percentage increases as the distance increased for each scenarios.

User channel with higher SNR will has high channel capacity or data rate. This channel

capacity also considered as the data rate achieved which will determine the scheduling

decisions.

Figure 6(a). Channel Capacity of Scenario A

Figure 6(b). Channel Capacity of Scenario B


Vol. 11, No. 12 (2018)


Figure 6(c). Channel Capacity of Scenario C

Figure 6(d). Channel Capacity of Scenario D

4.3 Scheduling

Proportional Fair algorithm prioritizes serving for user with the better channel condition

level than the average level. This simulation resulting the average number of resource

blocks allocated to the user with certain condition. There are 50 resource blocks in a 10

MHz bandwidth system that can be allocated to some active users. Tables below show the

results of scheduling that the average of resource blocks allocated are directly proportional

with the data rates of each users in a LOS and NLOS conditions.

Table 3. Resource Block Allocation of Scenario A

User

number

Data rate (Mbps) RB

LOS NLOS LOS NLOS

1 191.926 132.449 14 17

2 169.192 96.081 12 12

3 169.192 96.081 12 12

4 155.926 74.889 11 9

Source: Calculation


Vol. 11, No. 12 (2018)


Table 4. Resource Block Allocation of Scenario B

User

number

Data rate (Mbps) RB

LOS NLOS LOS NLOS

1 191,926 132,449 7 8

2 191,926 132,449 7 8

3 173,926 103,652 6 6

4 173,926 103,652 6 6

5 173,926 103,652 6 6

6 165,189 89,681 6 5

7 165,189 89,681 6 5

8 155,926 74,889 6 5

Source: Calculation

Table 5. Resource Block Allocation of Scenario C

User

number

Data rate (Mbps) RB

LOS NLOS LOS NLOS

1 191,926 132,449 5 5

2 191,926 132,449 5 5

3 191,926 132,449 5 5

4 176,662 108,029 4 4

5 176,662 108,029 4 4

6 176,662 108,029 4 4

7 165,189 89,681 4 4

8 165,189 89,681 4 4

9 165,189 89,681 4 4

10 155,926 74,889 4 3

11 155,926 74,889 4 3

12 155,926 74,889 3 2

Source: Calculation

Table 6. Resource Block Allocation of Scenario D

User

number

Data rate (Mbps) RB

LOS NLOS LOS NLOS

1 191,926 132,449 4 4

2 191,926 132,449 4 4

3 191,926 132,449 4 4

4 191,926 132,449 4 4

5 179,721 112,922 3 4

6 179,721 112,922 3 4

7 165,189 89,681 3 3

8 165,189 89,681 3 3

9 165,189 89,681 3 3

10 165,189 89,681 3 3

11 161,721 84,139 3 3

12 161,721 84,139 3 3

13 155,926 74,889 3 2

14 155,926 74,889 3 2

15 155,926 74,889 3 2

16 155,926 74,889 1 2

Source: Calculation


Vol. 11, No. 12 (2018)


4.4 BER

BER value will be calculated based on modulation types those are QPSK and 16QAM

as the property which BER value are less than 10-3 for each modulation type [12]. Table 7

shows the calculation result that 16QAM has a higher value than QPSK because 16QAM

modulates data with the more bits number than QPSK. It also directly proportional with the

SNR value, as the channel with higher SNR will be modulated by 16QAM modulation

scheme so it has higher BER value. While in the increasing of user number conditions,

system BER tend to constant. BER in LOS condition are less than the value in a NLOS

condition for both modulation scheme. So the channel in LOS condition has better

performance than channel in a NLOS condition.

Table 7. System BER Performance

Scenario

BER

QPSK 16QAM

LOS (*10-22) NLOS (*10-11) LOS (*10-9) NLOS (*10-4)

A 0.1092 0.6393 0.2981 0.1306

B 0.0637 0.3593 0.2193 0.0843

C 0.1214 0.6437 0.3213 0.1356

D 0.1296 0.5914 0.3660 0.1430

Source: Calculation

4.5 Throughput

Normalized throughputs which analyzed are directly proportional with the number of

average resource block allocation. Table 8 and Figure 7 show the maximum value of

normalized throughput reached at 1000 meters for each scenarios. These value are tend to

increase as the user number inceased because of the relativity to the average data rate value

of the constant resource blocks number and distance range. Throughput values in the Table

11 reach more than one because the data rates of user channel are above the average value.

It represents the best value can be reached by the system.

Table 8. Maximum Throughput

Scenario User number Throughput

LOS NLOS

A 4 1,1187 1,3261

B 8 1,1031 1,2765

C 12 1,1131 1,3080

D 16 1,1228 1,3394

Source: Calculation


Vol. 11, No. 12 (2018)


Figure 7. Throughput to User Variation

4.6 Fairness

Fairness value between 0-1 are determined by the data rate value. Fairness level of each

scenarios showed at Table 9 and Figure 8 which fairness level at a LOS conditions are

higher than NLOS and tend to constant to the variation of user numbers. Fairness value are

close to 1 and represent the Proportional Fair algorithm property that calculate the average

condition of system. The scheduling depends on channel condition will provide fairness

within the resource allocation in proportional number.

Table 9. Fairness Level

Scenario User number Fairness

LOS NLOS

A 4 0,9949 0,9618

B 8 0,9952 0,9664

C 12 0,9930 0,9580

D 16 0,9926 0,9525

Source: Calculation

Figure 8. Fairness Level


Vol. 11, No. 12 (2018)


5. Conclusions

The variation of users distance to the eNodeB have implication on the channel condition

shown with SNR and channel condition or data rate. Then the resource blocks allocation

tend to decrease as the distance increases but also relative to the LOS and NLOS condition

on the same user because of different value of average data rates and channel condition at

LOS or NLOS. The Proportional Fair algorithm prioritizes channels with the highest

channel capacity for the average channel capacity of a user on a system to serve. This

priority means that the user's channel with the best conditions will be allocated higher than

any other user regardless of queue or retransmission. Users with the same distance could

have different value of channel capacity at LOS and NLOS so it has different number of

resource blocks allocated. However the allocation could be relative to LOS and NLOS

conditions for users with the same position in the same scenario.

By giving treatment in the form of variations in the number of users on the scheduling

with different scenarios BER, throughput, and fairness will show different performances.

SNR gives effect to the type of modulation applied to uplink LTE communication that is

QPSK and 16QAM. Furthermore this type of modulation will determine BER. Channels

with smaller BER values will have better quality. The BER calculation is influenced by the

modulation numbers of the different modulation types. BER system and fairness level are

tend to constant to the variation of user number because the value of the system will depend

on the average value. Normalized throughput and fairness are directly affected by the data

rate. The more user number, will decrease the average value of throughput of each scenarios.

While the maximum values are constant which reached by user with the best condition. So

the maximum normalized throughput tend to increase as the user number increases, it can

be interpreted that the Proportional Fair algorithm has a throughput performance that

supports multiuser diversity.

References

[1] E. Yaacoub and Z. Dawy, “Centralized and Distributed LTE Uplink Scheduling in a Distribute Base

Station Scenario,” Advances in Computational Tools for Engineering Application, ACTEA, (2009).

[2] C. Cox, “An Introduction to LTE, LTE Advance, SAE, and 4G Mobile Communication,” Wiley, US,

(2012).

[3] A. Jalali, R. Padovani and R. Pankaj, “Data Throughput of CDMA-HDR a High Efficiency-High Data

Rate Personal Communication Wireless System,” Vehicular Technology Conference Proceeding, 2000.

IEEE 51st, vol.3, (2000), pp.1854-1858.

[4] J. Kim, D. Kim, Y. Han, “Proportional Fair Scheduling Algorithm for SC-FDMA in LTE Uplink,”

International Journal of Global Communication Conference (GLOBECOM), IEEE. (2012).

[5] M.I. Salman, M.Q. Abdulhasan, C.K. Ng, N.K. Noordin, A. Sali, B.M. Ali, “Radio Resource Management

(RRM) for Green 3GPP Long Term Evolution (LTE) Cellular Networks: Review, and Trade-Offs”, The

Institution of Electronics and Telecommunication Engineers (IETE) Technical Review, vol. 30, no. 3,

(2013), pp. 257-269.

[6] Y. Cai, J. Yu, Y. Xu, M. Cai, “A Comparison of Packet Scheduling Algorithms for OFDMA Systems”,

ICSPCS. Signal Proccessing and Communcation Systems, vol., no (2008), pp.1-5.

[7] M. Schwartz, ”Information, Transmission, Modulation, and Noise”, McGraw Hill, New York, (1994).

[8] H. Shinsuke and R. Prasad, ”Multicarrier Techniques for 4G Mobile Communication”, Artech House.

(2003).

[9] Y. Barayan and I. Kostanic, ”Performance Evaluation of Proportional Fair ness Scheduling in LTE,”

Procceding of the World Congress on Engineering and Computer Science Vol II (2013).

[10] -, “Evaluation Methodology Document” IEEE 802.16 C802.16m07/080r2. [Online] http://ieee802.org/16

(2012).

[11] R. Jain, D. Chiu, W. Hawe, ”A Quantitive Measure of Fairness And Discrimination for Resource

Allocation in Shared Computer System”, DEC Technical Report 301 (1984).

[12] R. P. R. Murti, A. Fahmi, G. Budiman, “Analisis Kinerja Gabungan Modulasi Adaptif dan Channel

Dependent Scheduling pada Teknologi OFDMA Arah Downlink”, Thesis of Telkom University, Bandung,

(2011).


Vol. 11, No. 12 (2018)


Authors

Endah Budi Purnomowati received B.Sc. and M.Sc. from ITS,

Surabaya Indonesia in 1982 and 1996, respectively. She is with at

Electrical Engineering, Brawijaya University Malang. Her areas of

interest is mobile communication. She is a reseacher at

Telecommunication Laboratory, Electrical Engineering, Brawijaya

University Malang.

Email: [email protected]

Faculty of Engineering, Departement of Electrical Engineering,

Brawijaya University, MT. Haryono st 167 Malang, East Java,

Indonesia.

Muthia Rahma received her B.Sc. in Electrical Engineering from

Brawijaya University, Indonesia, in 2018. From 2016 to 2018 she has

been working as a laboratory assistant of Telecommunication

Laboratory Brawijaya University. Her research and development

interests are mostly about wireless communication network included

mobile telecommunication.




Indonesia.

Rudy Yuwono, was born in blitar, june15, 1971. He received

Bachelot Degree from university of Brawijaya, Malang Indonesia in

1997 and Master Degree from University of Kassel, Germany in 2005

Curently, he is working at Electrical Engineering, University of

Brawijaya Malang as Lecturer and Researcher. His research interest

are Antena and Propagation, Microwave and Reasercher.




Indonesia.

mailto:[email protected]




Vol. 11, No. 12 (2018)


Date post:	15-Aug-2019
Category:	Documents
Upload:	truongphuc
View:	215 times
Download:	0 times

Analysis of Proportional Fair Scheduling Algorithm for...

Documents